SmartHeal-Agentic-AI / src /wound_analysis.py
SmartHeal's picture
Upload 33 files
185c377 verified
raw
history blame
27.5 kB
import logging
import cv2
import numpy as np
from PIL import Image
import json
from datetime import datetime
import os
# Try to import AI libraries (graceful fallback if not available)
try:
from transformers import pipeline
TRANSFORMERS_AVAILABLE = True
except ImportError:
TRANSFORMERS_AVAILABLE = False
logging.warning("Transformers not available")
try:
from ultralytics import YOLO
YOLO_AVAILABLE = True
except ImportError:
YOLO_AVAILABLE = False
logging.warning("YOLO not available")
try:
import tensorflow as tf
TF_AVAILABLE = True
except ImportError:
TF_AVAILABLE = False
logging.warning("TensorFlow not available")
class WoundAnalyzer:
"""AI-powered wound analysis system"""
def __init__(self, config):
"""Initialize wound analyzer with configuration"""
self.config = config
self.models_loaded = False
self.load_models()
def load_models(self):
"""Load AI models for wound analysis"""
try:
# Load models if libraries are available
if TRANSFORMERS_AVAILABLE:
try:
# Load a general image classification model
self.image_classifier = pipeline(
"image-classification",
model="google/vit-base-patch16-224",
token=self.config.HF_TOKEN
)
logging.info("✅ Image classification model loaded")
except Exception as e:
logging.warning(f"Could not load image classifier: {e}")
self.image_classifier = None
if YOLO_AVAILABLE:
try:
# Try to load YOLO model (will download if not present)
self.yolo_model = YOLO('yolov8n.pt')
logging.info("✅ YOLO model loaded")
except Exception as e:
logging.warning(f"Could not load YOLO model: {e}")
self.yolo_model = None
self.models_loaded = True
logging.info("✅ Wound analyzer initialized")
except Exception as e:
logging.error(f"Error loading models: {e}")
self.models_loaded = False
def analyze_wound(self, image, questionnaire_id):
"""Analyze wound image and return comprehensive results"""
start_time = datetime.now()
try:
if not image:
return self._create_error_result("No image provided")
# Get questionnaire data for context
questionnaire_data = self._get_questionnaire_data(questionnaire_id)
# Convert image to various formats for analysis
cv_image = self._pil_to_cv2(image)
np_image = np.array(image)
# Perform basic image analysis
basic_analysis = self._basic_image_analysis(cv_image, np_image)
# Perform AI analysis if models are available
ai_analysis = self._ai_image_analysis(image)
# Use enhanced AI processor if available
try:
from .ai_processor import AIProcessor
ai_processor = AIProcessor()
# Perform visual analysis using AI processor
visual_results = ai_processor.perform_visual_analysis(image)
# Query clinical guidelines
query = f"wound care {questionnaire_data.get('wound_location', '')} {questionnaire_data.get('diabetic_status', '')}"
guideline_context = ai_processor.query_guidelines(query)
# Generate comprehensive report
comprehensive_report = ai_processor.generate_final_report(
questionnaire_data, visual_results, guideline_context, image
)
# Merge AI processor results with basic analysis
ai_analysis.update({
'visual_analysis': visual_results,
'clinical_guidelines': guideline_context,
'comprehensive_report': comprehensive_report
})
logging.info("Enhanced AI analysis completed")
except Exception as e:
logging.warning(f"Enhanced AI processor not available: {e}")
# Combine results
analysis_result = self._combine_analysis_results(
basic_analysis,
ai_analysis,
questionnaire_id
)
# Calculate processing time
processing_time = (datetime.now() - start_time).total_seconds()
analysis_result['processing_time'] = processing_time
logging.info(f"Wound analysis completed in {processing_time:.2f} seconds")
return analysis_result
except Exception as e:
logging.error(f"Wound analysis error: {e}")
return self._create_error_result(f"Analysis failed: {str(e)}")
def _get_questionnaire_data(self, questionnaire_id):
"""Get questionnaire data for analysis context"""
try:
# This should connect to the database to get questionnaire data
# For now, return empty dict as fallback
return {}
except Exception as e:
logging.warning(f"Could not fetch questionnaire data: {e}")
return {}
def _pil_to_cv2(self, pil_image):
"""Convert PIL image to OpenCV format"""
try:
# Convert PIL to RGB if not already
if pil_image.mode != 'RGB':
pil_image = pil_image.convert('RGB')
# Convert to numpy array and then to OpenCV format
np_array = np.array(pil_image)
cv_image = cv2.cvtColor(np_array, cv2.COLOR_RGB2BGR)
return cv_image
except Exception as e:
logging.error(f"Error converting PIL to CV2: {e}")
return None
def _basic_image_analysis(self, cv_image, np_image):
"""Perform basic image analysis using OpenCV"""
try:
analysis = {}
if cv_image is not None:
# Image properties
height, width = cv_image.shape[:2]
analysis['dimensions'] = f"{width}x{height}"
analysis['image_quality'] = self._assess_image_quality(cv_image)
# Color analysis
analysis['color_analysis'] = self._analyze_colors(cv_image)
# Texture analysis
analysis['texture_analysis'] = self._analyze_texture(cv_image)
# Edge detection for wound boundaries
analysis['edge_analysis'] = self._analyze_edges(cv_image)
return analysis
except Exception as e:
logging.error(f"Basic image analysis error: {e}")
return {}
def _assess_image_quality(self, cv_image):
"""Assess image quality metrics"""
try:
# Calculate sharpness using Laplacian variance
gray = cv2.cvtColor(cv_image, cv2.COLOR_BGR2GRAY)
sharpness = cv2.Laplacian(gray, cv2.CV_64F).var()
# Calculate brightness
brightness = np.mean(cv_image)
# Calculate contrast
contrast = np.std(cv_image)
# Determine quality rating
if sharpness > 500 and 50 < brightness < 200 and contrast > 30:
quality = "Good"
elif sharpness > 100 and 30 < brightness < 230 and contrast > 15:
quality = "Fair"
else:
quality = "Poor"
return {
'sharpness': float(sharpness),
'brightness': float(brightness),
'contrast': float(contrast),
'overall_quality': quality
}
except Exception as e:
logging.error(f"Image quality assessment error: {e}")
return {'overall_quality': 'Unknown'}
def _analyze_colors(self, cv_image):
"""Analyze color properties of the wound"""
try:
# Convert to HSV for better color analysis
hsv = cv2.cvtColor(cv_image, cv2.COLOR_BGR2HSV)
# Calculate color statistics
mean_hue = np.mean(hsv[:, :, 0])
mean_saturation = np.mean(hsv[:, :, 1])
mean_value = np.mean(hsv[:, :, 2])
# Detect dominant colors
dominant_colors = self._get_dominant_colors(cv_image)
return {
'mean_hue': float(mean_hue),
'mean_saturation': float(mean_saturation),
'mean_value': float(mean_value),
'dominant_colors': dominant_colors
}
except Exception as e:
logging.error(f"Color analysis error: {e}")
return {}
def _get_dominant_colors(self, cv_image, k=3):
"""Get dominant colors in the image"""
try:
# Reshape image to be a list of pixels
data = cv_image.reshape((-1, 3))
data = np.float32(data)
# Apply k-means clustering
criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 20, 1.0)
_, labels, centers = cv2.kmeans(data, k, None, criteria, 10, cv2.KMEANS_RANDOM_CENTERS)
# Convert back to uint8 and get color names
centers = np.uint8(centers)
dominant_colors = []
for center in centers:
color_name = self._classify_color(center)
dominant_colors.append({
'rgb': center.tolist(),
'name': color_name
})
return dominant_colors
except Exception as e:
logging.error(f"Dominant colors error: {e}")
return []
def _classify_color(self, rgb_color):
"""Classify RGB color into medical color categories"""
r, g, b = rgb_color
# Simple color classification for wound assessment
if r > 150 and g < 100 and b < 100:
return "Red/Inflammatory"
elif r > 150 and g > 150 and b < 100:
return "Yellow/Exudate"
elif r < 100 and g < 100 and b < 100:
return "Dark/Necrotic"
elif r > 200 and g > 200 and b > 200:
return "White/Pale"
elif r > 100 and g > 50 and b < 100:
return "Pink/Healthy"
else:
return "Mixed/Other"
def _analyze_texture(self, cv_image):
"""Analyze texture properties"""
try:
gray = cv2.cvtColor(cv_image, cv2.COLOR_BGR2GRAY)
# Calculate Local Binary Pattern (simplified)
texture_variance = np.var(gray)
texture_mean = np.mean(gray)
# Determine texture category
if texture_variance > 1000:
texture_type = "Rough/Irregular"
elif texture_variance > 500:
texture_type = "Moderate"
else:
texture_type = "Smooth"
return {
'variance': float(texture_variance),
'mean': float(texture_mean),
'type': texture_type
}
except Exception as e:
logging.error(f"Texture analysis error: {e}")
return {}
def _analyze_edges(self, cv_image):
"""Analyze edges for wound boundary detection"""
try:
gray = cv2.cvtColor(cv_image, cv2.COLOR_BGR2GRAY)
# Apply Canny edge detection
edges = cv2.Canny(gray, 50, 150)
# Count edge pixels
edge_count = np.sum(edges > 0)
total_pixels = edges.shape[0] * edges.shape[1]
edge_ratio = edge_count / total_pixels
# Determine wound boundary clarity
if edge_ratio > 0.1:
boundary_clarity = "Well-defined"
elif edge_ratio > 0.05:
boundary_clarity = "Moderately-defined"
else:
boundary_clarity = "Poorly-defined"
return {
'edge_count': int(edge_count),
'edge_ratio': float(edge_ratio),
'boundary_clarity': boundary_clarity
}
except Exception as e:
logging.error(f"Edge analysis error: {e}")
return {}
def _ai_image_analysis(self, image):
"""Perform AI-based image analysis"""
try:
ai_results = {}
# Use image classifier if available
if hasattr(self, 'image_classifier') and self.image_classifier:
try:
classification = self.image_classifier(image)
ai_results['classification'] = classification[:3] # Top 3 results
except Exception as e:
logging.warning(f"Image classification failed: {e}")
# Use YOLO for object detection if available
if hasattr(self, 'yolo_model') and self.yolo_model:
try:
detection_results = self.yolo_model(image)
ai_results['object_detection'] = self._process_yolo_results(detection_results)
except Exception as e:
logging.warning(f"YOLO detection failed: {e}")
return ai_results
except Exception as e:
logging.error(f"AI image analysis error: {e}")
return {}
def _process_yolo_results(self, results):
"""Process YOLO detection results"""
try:
processed_results = []
for result in results:
if hasattr(result, 'boxes') and result.boxes:
for box in result.boxes:
processed_results.append({
'confidence': float(box.conf.item()) if hasattr(box, 'conf') else 0.0,
'class_name': result.names.get(int(box.cls.item()), 'unknown') if hasattr(box, 'cls') else 'unknown'
})
return processed_results
except Exception as e:
logging.error(f"YOLO results processing error: {e}")
return []
def _combine_analysis_results(self, basic_analysis, ai_analysis, questionnaire_id):
"""Combine all analysis results into a comprehensive report"""
try:
# Create comprehensive analysis result
result = {
'questionnaire_id': questionnaire_id,
'basic_analysis': basic_analysis,
'ai_analysis': ai_analysis,
'model_version': 'SmartHeal-v1.0'
}
# Generate summary
result['summary'] = self._generate_summary(basic_analysis, ai_analysis)
# Generate recommendations
result['recommendations'] = self._generate_recommendations(basic_analysis, ai_analysis)
# Calculate risk assessment
result['risk_assessment'] = self._calculate_risk_assessment(basic_analysis, ai_analysis)
result['risk_level'] = result['risk_assessment']['level']
result['risk_score'] = result['risk_assessment']['score']
# Determine wound type
result['wound_type'] = self._determine_wound_type(basic_analysis, ai_analysis)
# Extract wound dimensions
result['wound_dimensions'] = basic_analysis.get('dimensions', 'Unknown')
return result
except Exception as e:
logging.error(f"Results combination error: {e}")
return self._create_error_result("Failed to combine analysis results")
def _generate_summary(self, basic_analysis, ai_analysis):
"""Generate analysis summary"""
try:
summary_parts = []
# Image quality assessment
if 'image_quality' in basic_analysis:
quality = basic_analysis['image_quality'].get('overall_quality', 'Unknown')
summary_parts.append(f"Image quality: {quality}")
# Color analysis summary
if 'color_analysis' in basic_analysis and 'dominant_colors' in basic_analysis['color_analysis']:
colors = basic_analysis['color_analysis']['dominant_colors']
if colors:
color_names = [color['name'] for color in colors[:2]]
summary_parts.append(f"Dominant colors: {', '.join(color_names)}")
# Texture summary
if 'texture_analysis' in basic_analysis:
texture_type = basic_analysis['texture_analysis'].get('type', 'Unknown')
summary_parts.append(f"Texture: {texture_type}")
# Boundary clarity
if 'edge_analysis' in basic_analysis:
boundary = basic_analysis['edge_analysis'].get('boundary_clarity', 'Unknown')
summary_parts.append(f"Wound boundaries: {boundary}")
# AI classification summary
if 'classification' in ai_analysis and ai_analysis['classification']:
top_class = ai_analysis['classification'][0]
summary_parts.append(f"AI classification: {top_class.get('label', 'Unknown')}")
summary = "Wound Analysis Summary: " + "; ".join(summary_parts) if summary_parts else "Basic wound analysis completed."
return summary
except Exception as e:
logging.error(f"Summary generation error: {e}")
return "Wound analysis completed with limited information due to processing constraints."
def _generate_recommendations(self, basic_analysis, ai_analysis):
"""Generate treatment recommendations based on analysis"""
try:
recommendations = []
# Image quality recommendations
if 'image_quality' in basic_analysis:
quality = basic_analysis['image_quality'].get('overall_quality', 'Unknown')
if quality == 'Poor':
recommendations.append("Consider retaking the image with better lighting and focus for more accurate analysis.")
# Color-based recommendations
if 'color_analysis' in basic_analysis and 'dominant_colors' in basic_analysis['color_analysis']:
colors = basic_analysis['color_analysis']['dominant_colors']
for color in colors:
color_name = color.get('name', '')
if 'Red/Inflammatory' in color_name:
recommendations.append("Red coloration may indicate inflammation. Monitor for infection signs.")
elif 'Yellow/Exudate' in color_name:
recommendations.append("Yellow areas suggest possible exudate. Consider wound cleansing.")
elif 'Dark/Necrotic' in color_name:
recommendations.append("Dark areas may indicate necrotic tissue. Consult for debridement evaluation.")
elif 'Pink/Healthy' in color_name:
recommendations.append("Pink coloration suggests healthy granulation tissue - positive healing sign.")
# Texture-based recommendations
if 'texture_analysis' in basic_analysis:
texture_type = basic_analysis['texture_analysis'].get('type', '')
if 'Rough/Irregular' in texture_type:
recommendations.append("Irregular texture may require specialized wound care approach.")
# Boundary-based recommendations
if 'edge_analysis' in basic_analysis:
boundary = basic_analysis['edge_analysis'].get('boundary_clarity', '')
if 'Poorly-defined' in boundary:
recommendations.append("Poorly defined wound edges may indicate ongoing tissue breakdown.")
# General recommendations
recommendations.extend([
"Continue regular wound monitoring and documentation.",
"Maintain appropriate wound hygiene and dressing protocols.",
"Consult healthcare provider for persistent or worsening symptoms.",
"Follow established wound care guidelines for optimal healing."
])
return "; ".join(recommendations) if recommendations else "Standard wound care protocols recommended."
except Exception as e:
logging.error(f"Recommendations generation error: {e}")
return "Consult healthcare provider for appropriate wound care recommendations."
def _calculate_risk_assessment(self, basic_analysis, ai_analysis):
"""Calculate risk assessment based on analysis"""
try:
risk_score = 0
risk_factors = []
# Image quality factor
if 'image_quality' in basic_analysis:
quality = basic_analysis['image_quality'].get('overall_quality', 'Unknown')
if quality == 'Poor':
risk_score += 10
risk_factors.append("Poor image quality")
# Color-based risk factors
if 'color_analysis' in basic_analysis and 'dominant_colors' in basic_analysis['color_analysis']:
colors = basic_analysis['color_analysis']['dominant_colors']
for color in colors:
color_name = color.get('name', '')
if 'Dark/Necrotic' in color_name:
risk_score += 30
risk_factors.append("Possible necrotic tissue")
elif 'Red/Inflammatory' in color_name:
risk_score += 20
risk_factors.append("Signs of inflammation")
elif 'Yellow/Exudate' in color_name:
risk_score += 15
risk_factors.append("Possible exudate")
# Texture risk factors
if 'texture_analysis' in basic_analysis:
texture_type = basic_analysis['texture_analysis'].get('type', '')
if 'Rough/Irregular' in texture_type:
risk_score += 10
risk_factors.append("Irregular texture")
# Boundary risk factors
if 'edge_analysis' in basic_analysis:
boundary = basic_analysis['edge_analysis'].get('boundary_clarity', '')
if 'Poorly-defined' in boundary:
risk_score += 15
risk_factors.append("Poorly defined boundaries")
# Determine risk level
if risk_score >= 50:
risk_level = "High"
elif risk_score >= 25:
risk_level = "Moderate"
elif risk_score >= 10:
risk_level = "Low"
else:
risk_level = "Minimal"
return {
'score': min(risk_score, 100), # Cap at 100
'level': risk_level,
'factors': risk_factors
}
except Exception as e:
logging.error(f"Risk assessment error: {e}")
return {
'score': 0,
'level': 'Unknown',
'factors': ['Assessment error']
}
def _determine_wound_type(self, basic_analysis, ai_analysis):
"""Determine wound type based on analysis"""
try:
# This is a simplified wound type determination
# In a real system, this would use more sophisticated ML models
wound_characteristics = []
# Color-based characteristics
if 'color_analysis' in basic_analysis and 'dominant_colors' in basic_analysis['color_analysis']:
colors = basic_analysis['color_analysis']['dominant_colors']
for color in colors:
color_name = color.get('name', '')
if 'Red/Inflammatory' in color_name:
wound_characteristics.append("inflammatory")
elif 'Pink/Healthy' in color_name:
wound_characteristics.append("granulating")
elif 'Yellow/Exudate' in color_name:
wound_characteristics.append("exudative")
elif 'Dark/Necrotic' in color_name:
wound_characteristics.append("necrotic")
# Determine primary wound type
if "necrotic" in wound_characteristics:
return "Necrotic wound"
elif "inflammatory" in wound_characteristics and "exudative" in wound_characteristics:
return "Infected/Inflammatory wound"
elif "granulating" in wound_characteristics:
return "Healing/Granulating wound"
elif "exudative" in wound_characteristics:
return "Exudative wound"
else:
return "Acute wound"
except Exception as e:
logging.error(f"Wound type determination error: {e}")
return "Undetermined wound type"
def _create_error_result(self, error_message):
"""Create error result structure"""
return {
'error': True,
'summary': f"Analysis Error: {error_message}",
'recommendations': "Please ensure image quality is adequate and try again. Consult healthcare provider if issues persist.",
'risk_level': 'Unknown',
'risk_score': 0,
'wound_type': 'Unknown',
'wound_dimensions': 'Unknown',
'processing_time': 0.0,
'model_version': 'SmartHeal-v1.0'
}